This project is an upgrade and redesign of the previous twisted-pair winding machine designed to wind cable for use in Oscilloscope isolation transformers (see: Twisted-Pair Cable Winding Machine).
Background
At a previous workplace, we specialized in the production of high-voltage isolated oscilloscopes. Each scope features four independent 2kV isolated input channels, connected only by optic fiber links, and a specially designed low-noise isolation transformer (rated to 20kV!).
Each transformer relies on a set of twisted-pair windings, originally produced using a manual process that was slow, tedious, and most importantly, poorly repeatable, which had significant implications on the performance of the scope. This prompted the creation of the first winding machine, a purely analog device, which was able to automatically twist wire to one of eight pitch options, and cut it to any length up to 200mm.
While the machine performed effectively, several design flaws and limitations led to irreparable wear after around 500 meters of wire throughput. To replace it a new machine has been designed and built, with significant improvements over the previous iteration.
Problems with the V1 Winder
- Pinch roller wear – Because the old design used only a single set of rollers, the inevitable slippage of the wire between them eventually wore through the rubber.
- Unrepairable pinch rollers – the pinch rollers were mounted in 3D Printed plastic bushings, which were bonded directly onto the twisting armature, making them excessively difficult to replace.
- Crude tensioner design – the tensioner used a simple pair of friction washers to control wire tension, with an adjustable spring to set compression force. This design had only indirect control of wire tension via its friction coefficient, lowering the repeatability and reliability of the process.
- Poor reliability – The machine was prone to problems with the tensioner, and the unreliable hair trigger and solenoid clutch for the wire cutter.
- Lack of flexibility – pitch could only be selected from one of eight options (set by exchanging drive gears), wire length was a maximum of 200mm, limited by the need to place the wire-cut trigger at maximum extent of the output wire.
- Difficulty of wire use – as with the manual winding process, the ends of the twisted pair tended to separate during the transformer assembly process. This has a significant impact on the assembly time of each transformer.
Improved Winder Design
The new machine uses two stepper motors to achieve infinitely variable options for wire pitch, over the limited eight gear ratios of the previous design.
The use of steppers also means that pitch can be varied during operation, which allows for an additional feature of creating ‘hard twists’ at the wire ends, which significantly improve the ease of winding the transformers by preventing unravelling.
This is accomplished by altering the step ratios such that both steppers turn at the same speeds, causing the wire to be twisted without advancing. During this operation, a motor driven clamp pinches the wire a short distance from the twisting armature, fixing the length of the hard twist.
The tensioner has also been redesigned from the relatively primitive friction washers, to a closed-loop feed motor allowing a constant and balanced tension to be maintained.
Twisting Armature
The redesigned twisting armature has changed significantly from the original version. Notably:
- Two sets of rollers are now used, a rear capstan around which the incoming wire is wrapped once, and a front pinch roller pair. The ratios are such that the pinch rollers turn ~4% faster, ensuring that the wire remains tight on the capstan at all times. The function of this is to completely eliminate the wire slippage that occurred in the original design (see: the Capstan equation). The small amount of slippage necessary for the pinch rollers is tolerable, as they are not carrying the wire tension load.
- The assembly has been redesigned to use only 3D Printed parts, which meant substituting the original worm gear (difficult to print) for a bevel gear and compound reduction set, for a overall reduction ratio of 22.5:1. The gears and rollers run on 3mm stainless shafts, with the exception of the primary bevel gear and timing pulley (red), which rides on a pair of ‘6702’ bearings, and the secondary bevel gear (green), which uses a pair of small ‘693’ bearings salvaged from PC fans.
- The armature frame is a two-piece design which can be printed with minimal support in critical areas. Compared to the original design, the new frame is bolted together rather than glued, enabling full disassembly for maintenance or repair.
Tensioner
The new tensioner uses a buffer system, consisting of an input metering roller, tensioning idler, and output idler. The tensioning idler is pivoted on a spring loaded arm, allowing it to raise as the twisting armature pulls wire through. A microswitch on the arm connects to the metering roller motor, and will trigger the motor to refill the tensioning buffer once the arm reaches a certain height.
The tension can be set by shifting the tensioning spring, providing a near-constant force, particularly at higher spring preloads (spring force drops somewhat over the travel arc of the tensioning arm.
Output Handling
Twisted pair spins with the twisting armature as it exits, necessitating a module to hold it in place before cutting. This is the function of the output tray, which uses a split cylindrical channel, which constrains with movement laterally, but can be opened using a set of three solenoids to release it into an output tray when cut. The solenoids operate in pull-only, so a set of adjustable 3D Printed flexure springs are mounted to the top of the tray to provide closing force.
Interspersed along the output tray are small tabs, which ‘kick’ the wire out as it opens, to prevent the wire from getting caught in one side of the half-channel.
The output system proved quite difficult to get to work reliably in practice. This was mainly due to the flawed design of the wire cutter due to the shear orientation, which had a tendency to bend the wire during cutting, causing it to not enter the channel properly.
An attempt was made to solve this by adding a solenoid wire clamp, which constrains the wire during twisting, but opens during cutting to allow the wire to move with the cutter blade and prevent permanent bending. In the prototype version shown below, the solenoid was triggered by a simple microswitch fixed under the blade end. The clamp helped with the problem, however to make it fully reliable the cutter mount would likely have to be redesigned to change the blade orientation.
Cutter and Pinch
The cutter and shears were designed to share the same eccentric-crank actuation mechanism to save development time, with the only difference being the replacement of the shears with a pair of 3D Printed jaws for the pinch. The large gear which drives the eccentric has a number of holes around its circumference, which can be fitted with timing pins to adjust the timing of the device itself, and any additional mechanisms, such as the output tray.
Electronics and User Interface
The system is controlled by an Arduino Nano, with two DRV8825 stepper drivers, and an L293D H-Bridge IC to provide motor control. Power for the DC motors is provided by a 5V fixed-output buck converter, separate from the linearly-regulated 5V logic supply to minimize electrical interference.
A 5110 LCD module is used as the user interface. These relatively inexpensive modules were originally used in Nokia 5110 mobile phones, and have become a common staple for hobbyist electronics.
The UI features a Settings menu, where run-parameters can be edited and stored into non-volatile EEPROM, and additional subroutines run. Included is a manual twist, and manual advance mode, the latter allows wire to be driven through the twisting armature on command, something which on the original machine was a relatively difficult task due to the need to manually rotate the armature.
The original circuit was prototyped on Veroboard, which has now been replaced by a custom 2-layer PCB using SMD and through-hole components. This has allowed the board footprint to be reduced to just 100x100mm, and should prove much more reliable.
